Dr. Shlomit Wagman is the Chief Regulatory and Compliance Officer at Rapyd, a global fintech unicorn, and a Senior Fellow at Harvard University. She previously served as the CEO of Israel’s Money Laundering and Terror Financing Prohibition Authority, Acting Chair of the Israel Privacy Protection Authority, and Chair of the Operational Working Group at the Financial Action Task Force (FATF), the global standard-setter on illicit finance. A recognized global expert in fintech, cryptocurrencies, AI, financial crime, and counter-terrorism financing, Wagman advises governments, has testified before the U.S. Senate, and has been recognized multiple times as one of the most influential figures in the Israeli economy. She holds a Ph.D. and LL.M. from Yale Law School and an LL.B. and B.A. from the Hebrew University.
Wagman was a speaker on a panel about industry’s use of AI moderated by The Innovator’s Editor-in-Chief at the Sparks Innovation Festival in Israel in March. She separately agreed to be interviewed on the rise of AI agents in finance and what the industry needs to do to prepare.
Q: How is AI changing financial services?
SW: The rise of autonomous AI agents in payments is, in my view, a critical shift in the financial world. In the near term, AI agents will execute financial transactions without human involvement – authorizing, verifying, and optimizing payments based on pre-defined policies and real-time data. Imagine ordering a pizza from your car using a voice command, buying concert tickets the moment they go on sale, or purchasing stocks at targeted prices – all seamlessly managed by AI in the background.
The agents will also transform how banks, credit card companies, and payment providers operate – navigating payment options and optimizing routing across systems. These agents will operate across integrated infrastructures – banks, payment providers, credit cards, and crypto – enabling seamless, instant, and cost-efficient cross-border transfers.
This development also presents a compelling use case for stablecoins – often seen as a solution looking for a market. With AI agents and a preset policy, you could use smart contracts supported by stablecoins to underpin cross-border transfer.
From the consumer’s perspective, AI will significantly enhance the user experience and empower more informed, efficient, financial decisions. For example, consumers will be able to automatically identify the most cost-effective loans or credit options based on their financial profile or interact with chatbots that simplify complex financial information and support personalized planning. Agentic AI will be able to do all of this in the background, handling authentication and verification and optimize payment challenges, including different currencies and payment methods.
There will be much more transparency around fees that until now have been complex and opaque. To facilitate that, platforms will need to connect and a lot of work will need to be done behind the scenes to build this new ecosystem, such as putting into place the right regulatory safeguards, money laundering protections, interoperability mechanisms, and consumer protection frameworks.
Q: Will Agentic AI replace financial service employees?
SW: AI will replace many low skill employees. For example, at Rapyd we automated large parts of the onboarding processes using AI. This leads to a faster and more efficient onboarding process that provide clients with an immediate and seamless customer experience, with less resources invested. We drop all the documents the client submitted in one place and the AI extracts all the relevant information, verifies details, and checks that policies are being implemented correctly. Using AI for these kinds of low-level tasks can save companies time and money. Routine tasks that once took minutes or even hours can now be completed in seconds.
In addition, AI agents work 24/7, in multiple languages, they never complain, and they don’t require milk to be supplied to the office coffee station. They can also spot anomalies in data better than humans can. We had a case recently in which a customer submitted a passport and other documents, and the AI immediately noticed that the ID numbers were not the same. AI also implements policies more consistently and accurately than low skilled employees.
That said the AI needs to be closely monitored because if something goes wrong, it goes wrong across all services at scale. That is precisely why we need more highly skilled employees to analyze and inspect issues and problems.
We are going to see an inverted labor pyramid. It is going to shift from a wide base of low skilled employees who escalate cases up to narrower layers of managers to a far smaller number of low skilled employees and a large group of skilled senior team members who will be needed to address AI’s output, ensure its accuracy, and shape the policies the algorithms are trained on. The sector will additionally need experts in AI to do the prompting correctly.
The key takeaway is that AI will help companies scale up new types of services in a systematic, cost-effective way, eliminating low-level jobs but they will need more sophisticated experts who understand the big picture and can steer AI in the right direction.
Q: Last week The Innovator interviewed Lior Zalmanson, a senior lecturer (assistant professor) at the Coller School of Management, Tel Aviv University, where he leads AIMLAB (Artificial Intelligence in Management, Labor and Business). In a series of experiments he and his PhD student, Yotam Liel, found that a growing number of people accept nonsensical AI outputs without question. They’re learning to trust AI reflexively. And while AI often performs better than humans on specific tasks, that only deepens the dilemma: when is trust justified?
SW: This is a great point. It is why we need more skilled workers who are experts in their domain to evaluate AI’s output.
Q: Speaking of trust, how will AI-powered fraud, deepfake impersonations, and synthetic identity attacks impact the resilience of our financial systems and the public’s faith in them?
SW: Innovation introduces new risk. GenAI helps illicit actors to conduct phishing campaigns personalized to each user for literally no cost: they can simply ask AI agents to crawl data from LinkedIn to create a phishing campaign in seconds that is so convincing that people are going to click that link. A friend of mine was tricked by what she thought was an invitation to speak about her specialty at a high-profile conference. It is also becoming very easy to pretend to be someone else. GenAI can imitate voices that are used as account identification, and they can create very convincing fake passports and IDs and use them to access your bank accounts or create new ones by successfully passing traditional KYC processes.
The traditional toolboxes used to detect fraud are no longer effective. AI systems are better at detecting deep fakes than humans. They can see patterns in how your customers are being approached and attacked. To prepare and build trust the industry must invest in advanced technologies that identify deepfakes, distinguish machines from humans, detect anomalies, secure systems from adversarial AI, and comply with evolving regulatory frameworks.
Anti-fraud detection tools to combat deepfakes are relatively underfunded. We need to develop these tools and international standards to protect the economy and society as the technology can impact democratic process and national security. We need Western countries to join forces on preserving the safety of AI tools and the integrity of our financial systems. Last, but not least, we need the public to realize that every piece of information could be a deepfake. We need a call for action for the technology and finance sectors to address these issues. If it doesn’t all the revenue financial services companies earn from improvements to their business from AI will be lost to all the damage the technology can do.
Q: How can traditional financial services companies best prepare for Agentic AI?
SW: Companies need to find the right use cases to start with, as not every use case is good for AI. In addition, right now, the cost of using GenAI does not reflect the actual cost. Just like with Cloud services the cost will go up. Think about this from day one. Also, think about how you want employees to use AI. Do they rely on AI for everything they do? Not all processes at this stage are best resolved with AI. Many actions must still be reviewed by humans, and you will need experts in their domain to do that to avoid mistaken outputs. Most importantly – make sure you are protecting yourself and that employees are not channeling sensitive information into public AI services processing data. For example, it looks as if employees working on a draft court decision were consulting an AI engine as we were able to find information on the case – before it was officially published – by asking certain questions to a public AI.
I am convinced that traditional financial services companies and innovative fintechs are all going to adopt AI very quickly. It allows them to cut costs and become more efficient, which is crucial in financial services. Companies that don’t will not be able to compete. This is a chance for larger financial and fintech companies to make sure they stay relevant. Cost reductions and enhanced customer experience will be the first big wins. Traditional banks should embrace these changes as soon as possible and collaborate with other stakeholders to remain innovative and agile.
To access more of The Innovator’s Interview Of The Week articles click here.
